AI Governance Principles

Effective Date: January 1, 2026 Version: 1.0 Last Updated: March 1, 2026

Estimatics deploys artificial intelligence in a domain with meaningful consequences — property damage assessment, insurance claims, and legal proceedings. Our AI governance framework reflects the responsibility that entails.

1. Core Principles

1.1 Human Oversight

AI systems at Estimatics are designed to assist licensed professionals, not to replace their judgment. Every AI output — damage findings, scope items, report narratives, geometry measurements — is presented as a recommendation requiring human review. No AI output is submitted to any external party without explicit user action.

1.2 Transparency

We disclose which AI systems we use, for what purposes, and what their limitations are. See our AI & Automated Systems Disclosure for specific model and provider information.

1.3 Accuracy Over Speed

In damage assessment and insurance documentation, an incorrect output has real consequences — denied claims, underpayment, or liability. We prioritize model accuracy over processing speed and continuously evaluate model performance against ground truth data.

1.4 Auditability

On the CERTIFY plan, AI findings are included in a cryptographically certified evidence package. This creates a permanent, tamper-evident record of what the AI observed, when, and from which source data — enabling independent verification of AI-assisted documentation.

1.5 Continuous Evaluation

AI model performance is monitored continuously. We evaluate output quality against professional reviewer assessments and update model assignments when improved models become available.

2. Model Selection and Updates

2.1 Task-Based Routing

Estimatics operates a proprietary AI Provider Router that assigns the optimal model to each task based on defined capability requirements, accuracy needs, and cost efficiency. No single model is used for all tasks.

2.2 Model Updates

When frontier AI providers (OpenAI, Anthropic) release improved models, we update model assignments after internal evaluation. Configuration-level updates — changing which model handles which task — require no code changes and can be applied rapidly in response to new model availability or identified performance issues.

2.3 Provider Redundancy

The AI Provider Router supports automatic fallback to a secondary provider if the primary provider experiences an outage. This ensures continuity of AI-assisted features.

3. Data Practices

3.1 No Training on Customer Data

Customer Content — photographs, property data, scope items — is not used to train or fine-tune any AI model, by Estimatics or by any third-party provider, without explicit customer consent.

3.2 Minimal Data Transmission

We transmit only the data necessary to complete each AI task. For image analysis, this means the photograph and relevant context (area tag, property type). We do not transmit personal information about claimants or property owners to AI providers.

3.3 Sensitive Data Handling

Photographs submitted to the Platform may include personally identifiable features (vehicle license plates, visible individuals). We recommend that users avoid capturing unnecessary personally identifiable information in inspection photographs.

4. Bias and Fairness

4.1 Damage Detection Consistency

We are committed to ensuring that AI damage detection performs consistently across property types, geographic regions, construction styles, and damage categories. We monitor for systematic gaps in detection accuracy and address them through model updates and supplemental guidance.

4.2 No Discriminatory Use

The Platform may not be used to make decisions that discriminate on the basis of race, religion, national origin, sex, age, disability, or any other protected characteristic. Damage assessment must be based on documented physical evidence.

5. Accountability

The Estimatics engineering and product leadership team is accountable for the AI systems deployed in the Platform. Material changes to AI capabilities, providers, or governance are reviewed by leadership before deployment.

For questions or concerns about our AI governance practices: legal@aiestimatics.com